ANÁLISIS DE LOS RESULTADO

Whereas the previous experiments were designed to explore the relationship between immersion, presence, and memory using manipulations of immersion previously shown to enhance memory, Experiment Three employed a manipulation which has not reliably

demonstrated a benefit on episodic memory. Namely, this experiment featured a manipulation of the sophistication of environmental lighting effects in virtual environments. Previous studies assessing memory performance resulting from manipulations of lighting quality in Headset-VR have failed to show increased memory performance when comparing a realistically illuminated environment (High-Quality) with an environment featuring even lighting on all surfaces (Low- Quality; Mania et al., 2010; Mania, Robinson, & Brandt, 2005).

Although memory has not been shown to benefit from increased quality of lighting effects, might this property of immersion have an impact on a subject’s level of presence? Evidence on whether realistic illumination affects presence has been somewhat mixed. Mania and Robinson (2004) observed no significant variations in ratings of presence across three conditions which varied in the realism of lighting effects for a given virtual environment. In light of this finding, it is perhaps unsurprising that no effect of illumination quality on memory was observed in their subsequent research (e.g., Mania, Robinson, & Brandt, 2005) – if presence is indeed a factor which can mediate the relationship often observed between immersion and memory, then it is reasonable to assume that a property of immersion that fails to influence presence should likewise have no effect on memory performance. In contrast, other research has demonstrated that increasing the realism of lighting effects (e.g., the inclusion of shadows) does

indeed enhance a user’s rating of presence in a virtual environment (Slater et al., 2009; Khanna et al., 2006; see also Slater, Usoh, & Chrysanthou, 1995). These findings raise an alternative explanation concerning the relationship between presence and memory: perhaps this manipulation of immersion prompts an increase in presence in the absence of any effect on memory, suggesting that these two concepts may be independent in the context of certain manipulations of immersion. Indeed, this suggestion would be consistent with the outcome of Experiment Two, wherein increased immersion enhanced presence but had no effect on memory. Another possibility is that these same environments which enhanced presence via the inclusion of higher-quality lighting effects would have also enhanced memory; however, no memory assessment was included in these experiments.

As none of the aforementioned studies on the effect of lighting quality on presence included a measure of memory performance, it is unclear what form the relationship between presence and memory might take with regard to this manipulation of immersion. Therefore, the following experiment was created in order to disentangle these contrasting possibilities and further examine the relationship between immersion, presence, and memory.

Experiment Three – Methods

With the exception of the details noted below, the methods employed in Experiment Three were identical to Experiment One.

Participants. Subjects were 32 undergraduate students (25 female) from UNC Chapel Hill participating in exchange for course credit in introductory psychology.

Design. Lighting quality was the manipulated property of immersion and resulted in the creation of two conditions: High-Quality and Low-Quality lighting. Specifically, the High-

Quality condition featured lighting which caused shadows and variable illumination throughout the environment with relation to a central source of light placed on the ceiling (see Figures 9a and 10a). As such, this condition realistically simulates how light emanating from a fixed point would be expected to act in natural environments. Notably, the lighting effects in the High- Quality condition were the same as those incorporated in Experiments One and Two. In contrast, the Low-Quality condition was characterized by flat lighting, such that all surfaces in the

environment were equally illuminated (thus precluding the occurrence of shadows or a gradient of brightness relative to a specific source of illumination; see Figures 9b and 10b). The resulting effect was a condition which is visually unrealistic – indeed, a perfect real-world analogue of the visual characteristics portrayed in this environment would be impossible to create. Finally, and as with Experiment Two, FoV was the same between both of these conditions and was

equivalent to the High-FoV condition in Experiment One.

Experiment Three – Results

Measures of Presence. Subject ratings on the ITC-SOPI Composite measure of presence revealed no difference between the High-Quality (M = 3.182, SD = 0.693) and Low-Quality (M = 3.094, SD = 0.669) lighting conditions, t(31) = -1.289, p = 0.207. Likewise, no difference between High-Quality (M = 4.651, SD = 0.658) and Low-Quality (M = 4.602, SD = 0.696) conditions was detected by the MCQ, t(31) = -0.681, p = 0.501. However, while the ITC-

Composite and MCQ scores did not significantly vary based on immersion condition, they were positively correlated with one another, r(30) = 0.520, p = 0.002.

Free Recall. The average number of intrusions was low (M = 0.531, SD = 0.842). Surprisingly, the proportion of correct free recall responses was significantly lower in the High-

Quality condition (M = 0.569, SD = 0.152) than the Low-Quality (M = 0.650, SD = 0.196) condition, revealing a negative effect of immersion on item memory, t(31) = 2.304, p = 0.028. However, no significant relationship was found between overall free recall performance and presence as measured by the ITC-SOPI Composite score (r(30) = 0.200, p = 0.272), though a marginal trend was found on the MCQ (r(30) = 0.333, p = 0.062). Moreover, the difference in presence scores between immersion conditions (i.e., High-Quality minus Low-Quality presence values for each subject) did not correlate with a difference in free recall performance, regardless of whether the ITC-SOPI Composite score (r(30) = 0.203, p = 0.264) or the MCQ (r(30) = - 0.095, p = 0.605) was used as the measure of presence.

Recognition. Once again, the proportion of hits on the recognition test was high in both the High-Quality (M = 0.930, SD = 0.071) and Low-Quality (M = 0.944, SD = 0.073) conditions, and the false alarm rate was low (M = 0.041, SD = 0.049). Scores were converted to d’ values and compared between immersion conditions, revealing no significant difference in performance (t(31) = 1.274, p = 0.212), although, as before, the interpretability of this result is once again obscured by the presence of a ceiling effect.

Source memory performance (i.e., IO scores) in both the High-Quality (M = 0.872, SD = 0.128) and Low-Quality (M = 0.892, SD = 0.090) lighting conditions was significantly above chance (ps < 0.001). However, there was no significant difference in source memory

performance between immersion conditions, t(31) = 0.808, p = 0.426. Furthermore, there was no significant relationship between overall source memory and presence (ITC-SOPI Composite: r(30) = 0.083, p = 0.654; MCQ: r(30) = 0.243, p = 0.180), nor was there a correlation between the difference scores (High-Quality minus Low-Quality lighting conditions) of source memory and presence (ITC-SOPI Composite: r(30) = -0.268, p = 0.140; MCQ: r(30) = -0.136, p = 0.461).

Experiment Three – Discussion

As with Experiments One and Two, the results of this experiment once again failed to detect a significant correlation between difference scores (High- minus Low-Immersion) on presence and memory, and likewise found that source memory performance did not vary between immersion conditions. However, a surprising effect arose in free recall performance wherein memory for objects in the Low-Quality condition surpassed that of items observed in High-Quality lighting. This represents a notable deviation from the previous experiments in this study which failed to detect any effect (positive or negative) of immersion on memory, and will be explored further below. Nevertheless, in the absence of a relationship whereby differences in presence between immersion conditions was correlated with a change in memory performance, the preconditions for a mediation analysis were once again unmet in this experiment.

While null effects of lighting quality on memory would not be particularly unexpected given the previous literature on this property of immersion, it is not immediately clear why Low- Quality lighting would have produced superior free recall performance in this experiment. To investigate this issue, it is worth first considering how the perceptual characteristics of this immersion manipulation might directly influence performance. Prior research has demonstrated that even objects rendered in very rudimentary virtual environments (in which stimuli are

perceptually impoverished on a variety of dimensions, including lighting) can still be recognized by subjects (see Mourkoussis et al., 2010). Moreover, while lighting features like shadows have been found to be especially beneficial for perceiving spatial information (e.g., object distance and the relative locations of objects in space), they do not appear to be especially informative with regard to perceiving the shape and border of the objects themselves (Mamassian, Knill, &

Kersten, 1998), nor does the absence of shadows appear to prevent subjects from accurately recognizing the identity of an object (Braje, Legge, & Kersten, 2000). Thus, it is perhaps the case that perceptual information associated with lighting quality is fairly superfluous with regard to item-specific processing.

While the perceptual characteristics of the objects themselves may not have enhanced item memory, is it possible that the difficulty of recognizing objects may have varied between the Low- and High-Quality lighting conditions? Specifically, it seems plausible that perceptual fluency was comparatively diminished in the Low-Quality condition, which (perhaps

counterintuitively) might have ultimately benefited memory performance for stimuli in that condition. The concept of desirable difficulty (Bjork, 1994) suggests that some conditions which increase the difficulty of encoding may prompt more effortful processing that actually enhances retrieval performance. In many cases, memory for perceptually disfluent stimuli is enhanced relative to more fluent stimuli (although this finding is not universal; for discussion, see Yue, Castel, & Bjork, 2013). Could this process be driving the current results? Closer inspection suggests this is likely not the case. First, this effect appears to be restricted to mixed-list designs (see Susser, Mulligan, & Besken, 2013), while the current experiment more closely reflects a pure list design. Additionally, it is ambiguous whether the manipulation of lighting effects is similar enough to common manipulations of perceptual degradation found in this body of

literature (e.g., blurred vs. clear words), thus obscuring whether a meaningful comparison can be drawn. Finally, basic perceptual research suggests that the speed and accuracy of object

recognition do not appear necessarily influenced by the presence or absence of shadows (see Braje, Legge, & Kersten, 2000), at least so long as the cast shadow is congruent with the shape of the object and the direction of ambient illumination in the environment (Castiello, 2001; for

discussion, see Dee & Santos, 2011). Therefore, although promising at first glance, this explanation does not appear to hold up to scrutiny when applied as an interpretation of the current results.

In consideration of these findings, perhaps one should not expect that the purely perceptual aspects of the High-Quality lighting condition should inherently confer a benefit in item memory performance when compared to a Low-Quality condition. Could the same be said if a spatial memory task had been employed in the current study (given the aforementioned benefit of spatial perception conferred by the presence of realistic lighting)? Mania et al. (2010) did not observe an overall difference in spatial memory performance between levels of lighting quality in Headset-VR. However, subjects did indicate higher levels of confidence in

remembering the location of objects studied in the Low-Quality condition, suggesting a

metacognitive bias in favor of more basic lighting effects despite no actual difference in source memory accuracy. In contrast, when Mania, Robinson, and Brandt (2005) measured confidence ratings in a similar experiment on item memory, no significant difference in confidence ratings emerged between the lighting conditions, while the pattern of performance in object recognition was unclear (Mid-Quality lighting was significantly better than Low-Quality, but High-Quality was no better than either Mid- or Low-Quality). While the relationship between lighting quality and memory remains unclear, these findings raise the possibility that metamemory judgments of virtual environments may be sensitive to the specific nature of the retrieval task. As such, future research may stand to benefit from the continued inclusion of metamemory assessments to further clarify its relationship with performance on various retrieval tasks between immersion conditions. Nevertheless, a clear explanation of the free recall results in the current experiment

does not seem to emerge from an account of the perceptual or metamnemonic properties of lighting quality detailed in prior research.

An alternative explanation of the results obtained in this experiment lies in a

consideration of the extent to which the appearance of each virtual environment deviated from what could be expected in real life. In other words, it is worth considering if the Low-Quality condition was more visually distinctive than the High-Quality condition (by virtue of its

increased perceptual deviation from reality) and whether this distinctiveness may have resulted in improved memory performance. Distinctiveness effects take on several forms and, in many cases, result in enhanced memory performance for items which have peculiar or atypical characteristics. Many individual manipulations of distinctiveness fit into one of two broad categories: primary distinctiveness (when an item is distinct with regard to its immediate context) and secondary distinctiveness (when stimuli are unusual in absolute terms based upon prior knowledge, regardless of the specific context in which they are observed; see Schmidt, 1991). As such, to the extent that stimuli in the Low-Quality condition were distinct, it would be categorized as a form of secondary distinctiveness.

It is not entirely clear whether the occurrence of secondary distinctiveness can adequately account for the current results. Many secondary distinctiveness effects are most likely to emerge in recognition assessments of stimuli studied in a mixed-list design (Schmidt, 1991) – in contrast, the current results found an effect on free recall from stimuli studied in a manner more akin to a pure-list design. However, bizarreness effects (a subclassification of secondary distinctiveness) are a noteworthy exception for which effects are more pronounced in free recall (for review of bizarreness effects on memory, see Worthen, 2006). Additionally, there is evidence suggesting that bizarreness may be beneficial for item memory while having no effect on source memory

(e.g., Macklin & McDaniel, 2010). Moreover, mixed-list retrieval tasks like the one incorporated in this study (wherein recall for multiple sets of stimuli is evaluated simultaneously as opposed to after each individual study list) have produced bizarreness effects, even if the study phases feature a pure-list design (McDaniel, Dornburg, & Guynn, 2005). Finally, although more conventional paradigms (e.g., lists of sentences) are typically employed to study bizarreness effects, mnemonic benefits have also been observed when assessing memory for individual components of more comprehensive bizarre events (see Worthen, 2006).

In light of these observed properties of bizarreness effects, it appears as though a distinctiveness-based explanation of the current results provides a plausible account for the enhanced free recall performance in the Low-Quality condition. Future research would benefit from a more direct examination of this bizarreness account. Additionally, subsequent exploration of this topic should also seek to identify which properties of immersion result in an experience of bizarreness when manipulated. Indeed, while this account is consistent with Experiment Three, it does not apply to the results of Experiments One or Two (perhaps because the previous

experiments did not alter the visual appearance of the objects, but instead varied with regard to the visual boundaries of the environment and the number of sensory modalities expressed).

Regardless of the ambiguity in interpreting the results of the free recall assessment, the outcome of presence scores in this experiment is worth briefly discussing. In particular, a significant positive correlation in overall presence scores was found between the ITC-SOPI Composite and MCQ. This result contrasts with the outcomes of Experiment One (which only detected a marginal trend) and Experiment Two. This provides evidence of a clear positive relationship between these two metrics, and lends credence to the notion that the MCQ may be a valuable tool to study further in relation to presence in VR research. Additionally, neither

presence measurement detected a difference in overall presence between immersion conditions, suggesting that manipulating immersion via lighting quality did not strongly influence a user’s sense of being mentally transported to the virtual environment in this experiment. Finally, a marginal positive trend was observed between MCQ and overall free recall performance, although no such trend was detected for the ITC-SOPI composite. While it is worth noting that this pattern was directionally consistent with the results from Experiment One (which detected a positive relationship between overall free recall and both measures of presence), its failure to meet the significance threshold precludes any more substantive interpretation of this relationship in the current experiment.